LED Drivers for Lighting

The driver decides more about a finished luminaire than the LED package does. It sets the current the emitters see and therefore the light output, it determines how the fitting behaves when dimmed, it is responsible for the flicker performance, and it is almost always the component that fails first — LED packages are commonly rated for tens of thousands of hours, and in the field the electronics are what ends the fitting's life.

Specifying one starts with a question that has only two answers but is still got wrong regularly: does the load want constant current or constant voltage?

Constant current or constant voltage?

A constant-current driver holds the output current at a set value and lets the voltage settle wherever the load's forward voltage puts it. It is what LED modules, COB arrays and engines specified at a fixed current require, because light output tracks current and because an LED's forward voltage falls as it warms — drive one from a fixed voltage and current rises as it heats, which raises temperature, which raises current again.

A constant-voltage driver holds 12 V or 24 V and lets the load draw what it needs. LED strip and tape are built this way, with resistors integrated into each cuttable segment, so what the installation needs is a stable voltage rail rather than a controlled current.

Connecting these the wrong way round does not produce a subtle problem. A constant-current driver into a strip designed for constant voltage will drive it far outside its rating, and a constant-voltage driver into a bare LED array with no current limiting has no mechanism to stop the thermal runaway described above. The LED driver category carries both types, and several families — XLG, HLG and ELG among them — are available in constant-current, constant-voltage and combined variants that are distinguished only by a suffix. Reading the full part number matters here more than in most categories.

A third behaviour sits between them. Constant-power drivers hold output power constant across a wide voltage window, letting the current adjust as the load's forward voltage varies. For a manufacturer building one luminaire around several LED configurations, that reduces the number of driver variants to stock.

Flicker is a specification, not a matter of opinion

All mains-powered LED lighting modulates to some degree, because the supply itself is alternating. Whether that modulation matters is a question with a published answer: IEEE Std 1789-2015 gives recommended limits relating percent flicker to flicker frequency.

For the low-risk level the recommendation is that percent flicker stays below 0.025 times the frequency below 90 Hz, and below 0.08 times the frequency between 90 Hz and 1,250 Hz. Above 1,250 Hz the standard sets no limit. In practice that means a fitting flickering at 120 Hz — twice the line frequency, the usual case for a rectified mains supply — should stay under 10% modulation. A stricter no-observable-effect level applies 0.0333 times the frequency between 90 Hz and 3,000 Hz for sensitive populations. Separately, and independently of the other effects, percent flicker below 90 Hz should stay under 5% to avoid photosensitive seizure risk.

These numbers earn their place in a specification for offices, schools, healthcare and industrial spaces with rotating machinery, where a stroboscopic effect is a safety issue rather than a comfort one. They matter for any space where video is recorded, because a camera's shutter interacts with the modulation to produce banding that the eye never sees.

The part most often missed is that flicker performance changes with dimming. A driver that measures well at full output can perform very differently at 10%, particularly if it dims by pulse-width modulation. If the installation will be dimmed, the figure that matters is the one at the dimmed setting.

Dimming: match the protocol before anything else

A driver that cannot be commanded by the control system is the wrong driver regardless of its other merits. The mainstream options behave quite differently:

• 0–10 V and 1–10 V analogue — simple, robust, long-established, and requires a control pair to every fitting. Common on the D and B suffixed variants in the LPF, NPF and HLG families.
• DALI and DALI-2 — digital, addressable, bidirectional, so a fitting can report status back. DALI-2 adds certification of the control devices as well as the drivers. The DA2 suffix marks these variants in families such as XLG.
• PWM — used mainly for constant-voltage strip, where the supply itself is modulated. The frequency is what determines the flicker behaviour described above.
• Mains dimming — trailing or leading edge, using existing wall dimmers. Attractive on refurbishment because it needs no new cable, and the least predictable, because the driver has to cooperate with a specific dimmer. Compatibility is established by testing the combination, not by reading two datasheets.

Dimming range is a separate specification from protocol. A driver rated to 10% and one rated to 1% both accept the same command; only one of them will take a restaurant down to the level the designer drew. Where dimming to very low levels is part of the brief, it belongs in the driver selection rather than in the commissioning.

Matching the output window to the load

A constant-current driver produces its rated current only within a stated output voltage range. The load's forward voltage — the sum of the forward voltages of the LEDs in series — has to sit inside that window, and it has to stay there under conditions other than the bench.

Forward voltage moves. It falls as the junction warms, so a string measured cold sits higher than the same string at working temperature. It varies between production batches within the manufacturer's tolerance. It drifts slowly over the life of the LEDs. A design that places the string at the very edge of the driver's window will work on the prototype and fail on some fraction of production units.

Leaving margin at both ends of the window is the whole of the technique. Where a manufacturer is building the same fitting with different LED counts, the constant-power families with their wide output windows are what avoid a separate driver part number for each variant.

Outdoor, wet and exposed installations

Outdoor lighting subjects the driver to water, temperature cycling, ultraviolet exposure and surges induced on long exterior cable runs.

Ingress protection needs to reflect the actual exposure rather than the general idea of being outside — protection against jets is a different test from protection against immersion, and a driver in a below-grade enclosure that can flood needs the latter. Potted and encapsulated drivers, which most of the outdoor-rated families are, resist both moisture and vibration far better than open-frame construction, at the cost of being unrepairable and harder to cool.

Surge withstand is the specification that decides survival on exposed sites. A lighting circuit running across a car park or along a street picks up induced surges from nearby lightning even without a direct strike, and drivers are offered with different surge ratings for exactly this reason. Where the installation is exposed, the higher-rated variant is much cheaper than the access equipment needed to replace failed drivers on columns.

The LED accessory category covers the surrounding parts — dimming interfaces, timers and controllers — that turn a driver into a controlled fitting.

Efficiency, power factor and inrush

Efficiency determines how much of the input becomes light rather than heat in the driver, and the heat matters twice: it is energy wasted, and it raises the temperature of the electrolytic capacitors that usually determine the driver's service life.

Power factor describes how much of the apparent current drawn does useful work. Commercial installations are commonly required to meet a minimum figure, and a poor one means the building's wiring carries current that the meter does not bill for but the cable still has to be sized for.

Inrush is the specification that catches large installations. Every driver charges its input capacitance at switch-on, and while one driver's inrush is trivial, a hundred fittings on one circuit energising together produce a peak that will trip the breaker. The usual answers are staged switching across several contactors, breakers with a characteristic chosen to tolerate the surge, or distributing the fittings across more circuits. It is a calculation worth doing at the design stage, because the alternative is discovered at handover.